Microsoft's Surface Laptop Ultra launched October 16, following preorders that opened October 7 at a $2,599 starting price. The laptop is built on Nvidia's RTX Spark platform and targets developers, creators, and users running demanding local AI workloads, according to Microsoft's launch announcement and Tom's Hardware.
The launch coincided with new Windows 11 capabilities for agentic AI, including Copilot access to local files, support for local models, Windows ML for running models across GPU, NPU, and CPU hardware, and Microsoft Execution Containers to limit what AI agents can access. Engadget characterized the early benefits as most relevant to developers and power users, not a mainstream productivity upgrade.
The laptop has reached its announced release date, but several of the most prominent Copilot agent features Microsoft demonstrated are still planned for a staged rollout. Microsoft's launch materials also combine capacity, peak compute, and application demos, which measure different things, so it's worth separating what's confirmed from what's still a company claim.
Surface Laptop Ultra price and release date
The Surface Laptop Ultra starts at $2,599, with availability beginning October 16 alongside RTX Spark laptops from Asus, Dell, Lenovo, MSI, and other manufacturers. Microsoft also opened preorders for the Surface RTX Spark Dev Box, a workstation-oriented machine starting at $5,999, with shipments scheduled for November.
The launch extended beyond the Surface lineup. Nvidia also announced Windows support for DGX Spark and DGX Station, expanding the platform's reach across laptops and workstation-class systems.
The $2,599 entry configuration includes an 18-core Arm CPU, a Blackwell GPU with 5,120 CUDA cores, 24GB of unified memory, and a 512GB SSD, according to The Verge's pricing report. Those specifications are substantially below the maximum configuration Microsoft emphasizes in its marketing.
Higher-end models offer up to a 20-core Arm CPU, a Blackwell GPU with 6,144 CUDA cores, and 128GB of unified LPDDR5X memory. The laptop also features a 15-inch mini-LED display rated at up to 2,000 nits of peak HDR brightness and a magnetic USB-C charging connector Microsoft calls Magnetic Connect.
The top configuration reaches approximately $5,899, so anyone buying specifically for the maximum AI performance or memory capacity should check the exact configuration rather than assuming those specifications come with the base model.
What Microsoft's Surface Laptop Ultra AI performance claims actually establish
The 128GB memory pool is driving much of the attention because it's shared between the CPU and GPU rather than reserved for graphics alone. Microsoft says that capacity allows the system to run AI models exceeding 120 billion parameters locally.
That's a capacity claim. It describes the size of models the system can accommodate under suitable conditions, not how quickly the laptop actually runs them.
Microsoft also cites up to one petaflop of AI compute on the highest RTX Spark configuration. That number represents peak theoretical AI throughput under particular numerical precision and processing conditions, not sustained performance across everyday applications.
Separately, Microsoft compared the Surface Laptop Ultra against a MacBook Pro running Apple's M5 Pro chip, claiming up to 2.1 times faster time to first token, 4.3 times faster image generation, and 6.2 times faster video generation, according to Tom's Hardware.
Those figures come from Microsoft's own demonstrations and should not be treated as general performance advantages across all AI workloads.
Independent testing has since added useful context. Wired's review found strong local AI capabilities but less impressive conventional CPU performance, while The Verge's review highlighted the machine's AI and gaming strengths alongside limitations in traditional computing workloads.
A 128GB unified memory pool remains a meaningful capacity advantage for developers working with large models. Still, loading a 120-billion-parameter model and running it at usable, interactive speed are different problems. Quantization level, context length, memory bandwidth, and sustained thermal performance all affect the second one.
Copilot's new agent features: what's live now, and what's still cloud-dependent
Microsoft's Jacob Andreou demonstrated Copilot handling tasks such as searching for files, renaming and zipping documents, and drafting an email for approval, using a combination of local and cloud models. In one example, Copilot gathered tax documents for an accountant and prepared an outgoing email for review, according to Tom's Hardware.
Andreou said those agentic features would arrive over the following months, rather than necessarily being available when the laptop launched October 16. Buyers interested specifically in those capabilities should expect a staged software rollout.
Local processing doesn't mean every task runs offline, either. Microsoft's approach combines local and cloud models, with cloud processing continuing to play a role in many Copilot tasks. Buying an RTX Spark laptop doesn't automatically mean all AI processing stays on the device.
The permission system around these agents is also evolving. Microsoft Execution Containers are intended to restrict what agents can access, and Microsoft's Divya Venkataramu demonstrated a scenario in which agents had more limited file access than the user account running them.
The demonstration doesn't establish the exact default permissions or consent prompts every user will encounter across future agent features.
The underlying platform extends beyond this laptop. Windows ML provides a runtime for deploying models across supported GPU, NPU, and CPU hardware, while integration with tools such as llama.cpp is intended to give developers more local model options.
One branding detail stands out. Despite reportedly meeting the hardware requirements associated with Copilot+ PCs, Microsoft didn't emphasize that label during the launch. Windows Latest linked the omission to criticism surrounding Recall and Microsoft's broader Copilot strategy, though that's the outlet's interpretation rather than a confirmed explanation from Microsoft.
Who should buy a Microsoft Surface Laptop Ultra?
The 128GB configuration is one of the clearest reasons developers are watching the device. Aidan Pinero, an AI researcher and master's student at UC Santa Cruz Silicon Valley, told Engadget he'd been jealous of friends running 128GB MacBook Pros and welcomed having similar memory flexibility on a Windows machine.
Gamers willing to accept some early Arm compatibility risks also have reason for cautious optimism. Early hands-on demonstrations showed demanding games running on RTX Spark hardware, with ray tracing and Nvidia's graphics technologies helping performance in supported titles.
However, demonstrations are not a substitute for broad compatibility testing. Microsoft and Nvidia have highlighted hundreds of supported games, but individual titles, anti-cheat systems, and drivers still need to be checked.
Anyone coming from an x86 Windows machine should treat Arm compatibility as a buying consideration rather than an assumption. Some applications and games run natively, while others rely on Windows' Prism emulation layer, which can affect performance, compatibility, and battery life.
Independent reviews have begun answering those questions. Wired found that the Surface Laptop Ultra's conventional CPU performance lagged some competitors and that software compatibility remained a concern. The Verge reported battery life of roughly seven to 10 hours depending on workload, below what some buyers might expect from a premium Arm laptop.
For everyday productivity buyers, Engadget's original framing remains useful: the benefits are concentrated in developer, creative, and local AI workloads rather than a universal upgrade for typical office tasks.
The Surface Laptop Ultra makes the strongest case for buyers who specifically need large unified memory configurations, CUDA-based local AI development, or Nvidia's graphics ecosystem in a portable Windows machine. The $2,599 entry model is a different proposition from the 128GB flagship, and buyers should compare the exact configuration against their workload before ordering.
Everyone else should weigh independent battery, CPU, gaming, and app-compatibility results before switching away from an established x86 laptop. Microsoft's AI performance claims are promising, but they don't eliminate the trade-offs of a new Arm-based platform.
Comments
Be the first, drop a comment!